Research map / Graph learning / GNN architectures
Going Deeper into Permutation-Sensitive Graph Neural Networks
Zhongyu Huang, Yingheng Wang, Chaozhuo Li, Huiguang He
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use permutation-group sampling to model pairwise correlations among neighbors | PG-GNN · Incidence triangle counting, Erdős-Rényi random graph | MAE: 0.023 $\pm$ 0.011 -> 0.019 $\pm$ 0.002 | Improved |
| Use permutation-group sampling to model pairwise correlations among neighbors | PG-GNN · Incidence triangle counting, random regular graph | MAE: 0.037 $\pm$ 0.019 -> 0.027 $\pm$ 0.001 | Improved |
| Use permutation-sensitive aggregation instead of permutation-invariant aggregation | GNNs on substructure-counting tasks · Incidence triangle and 4-clique counting | MAE | Improved |
| Use PG-GNN rather than the strongest listed baseline | Graph classification · TUDataset social-network benchmarks | accuracy | Improved |
| Use PG-GNN rather than the strongest listed baseline | Graph classification · NCI1 | accuracy: 86.0 $\pm$ 1.8 -> 82.8 $\pm$ 1.3 | Hurt |
| Use PG-GNN rather than the strongest listed baseline | Graph regression · ZINC | MAE: 0.223 $\pm$ 0.008 -> 0.282 $\pm$ 0.011 | Hurt |
| Use PG-GNN rather than PNA | Graph classification · MNIST | accuracy: 97.41 $\pm$ 0.16 -> 97.51 $\pm$ 0.07 | Improved |
| Use PG-GNN rather than Deep-LRP | Graph regression · ZINC | Time / Epoch: 72s -> 6.92s | Hurt |
| Use one-hop rather than extended multi-hop neighborhoods | PG-GNN on molecular graphs · NCI1 and ZINC | performance | Mixed |
Closest papers by meaning
- On the Expressivity and Sample Complexity of Node-Individualized Graph Neural Networks (NeurIPS 2024 · similarity 0.89)
- Commute Graph Neural Networks (ICML 2025 · similarity 0.88)
- Covered Forest: Fine-grained generalization analysis of graph neural networks (ICML 2025 · similarity 0.88)
- REVISITING MULTI-PERMUTATION EQUIVARIANCE THROUGH THE LENS OF IRREDUCIBLE REPRESENTATIONS (ICLR 2025 · similarity 0.87)
- Homomorphism Expressivity of Spectral Invariant Graph Neural Networks (ICLR 2025 · similarity 0.87)
- Equivariance Everywhere All At Once: A Recipe for Graph Foundation Models (NeurIPS 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
